The purpose of this study was to test the hypothesis that being praised for prosocial behavior would be longitudinally associated with decreased depressive symptoms in adolescents. We adopted 10–24 years as the definition of the age of adolescence according to the recent review (10), thus called the study participants as early adolescents. This study used data from the Tokyo Teen Cohort study (TTC), a prospective population-based cohort study that is currently underway and aims to investigate the developmental trajectory of adolescents (11). Children born between September 2002 and August 2004 in three local governments in Tokyo (Setagaya, Mitaka, and Chofu) were randomly extracted using the Basic Resident Register. Of the 10,234 pairs of children and their primary caregivers who were asked to participate, 4,478 pairs agreed to cooperate in the baseline survey (the 1st wave of data collection) at the age of 10 years. An oversampling method was used with the goal of having 3,000 pairs remain in the 2nd wave survey (11). Given the low follow-up rate of low-income families, all 620 pairs whose annual household incomes was lower than five million yen were asked to participate in the 2nd wave survey. Of the remaining 3,858 pairs, 2,551 were randomly asked to participate in the 2nd wave. A total of 3,172 pairs were asked to participate in the longitudinal cohort study, resulting in 3,007 pairs participating in the second wave of data collection at the age of 12 years (follow-up rate was 94.8%).

In each of the 1st and 2nd waves of data collection, trained investigators visited the participants' home twice and interviewed children and their primary caregivers. At the first visit, the investigators explained the research to both the child and the primary caregiver, obtained written consent, and asked them to complete the self-report questionnaire by the second visit. At the second visit, the child and the primary caregiver separately answered a self-report questionnaire containing sensitive content and sealed it immediately after completion. All questionnaires and data were collected anonymously. At this visit, the investigators also administered face-to-face psychological tests to the child.

TTC is a joint study of three institutions (Tokyo Metropolitan Institute of Medical Science, the University of Tokyo, and the Graduate University for Advanced Studies) and has been approved by the ethics committees of the three institutions.

The children answered self-report questionnaires including items on experiences of being praised, depressive symptoms and other variables, such as the number of siblings. Caregivers answered self-report questionnaires that included questions about the caregiver's age, psychological distress, educational background and annual household income.

In the self-report questionnaire at the 1st wave, 10 year-old children were asked to freely describe answers to the question “What are you praised for?”. We dichotomized the answers to “prosocial praise group” and “other praise group.” We did not score the degree of prosociality but just divided the answers to the two groups. A qualitative classification was made by several researchers (DN, NT, HN, MT) as to whether prosocial behavior was included in the answers, double-checked by other researcher (DN, NT), and was finally confirmed by several experienced researchers (MM, TK). Based on a previous study (12), we defined prosocial behavior as voluntary, intentional behavior that results in benefits for another; the motive is unspecified and may be positive, negative, or both. Only children who clearly answered that they were praised for their prosocial behavior such as “helping with housework” were designated the “prosocial praise group,” and other children who did answer but did not include prosocial behavior in their responses such as “getting a good score on the exam” were designated the “other praise group.” Children who made multiple responses were also classified as the “prosocial praise group” if more than one of their answers included prosocial behaviors. Blank fields were treated as missing values.

There may be children who behave prosocially but not be praised for their prosocial behaviors. There may also be children who are praised for their prosocial behaviors but do not recognize they are praised for their prosocial behaviors. The children's recognition of being praised for prosocial behavior does not necessarily correspond to the children's objective prosocial behavior. Objective prosocial behavior of the 10 year-old children was assessed by the Strength and Difficulty Questionnaire (SDQ), for which the primary caregivers answered in the self-report questionnaire in the 1st wave of data collection. The subscale score for prosocial behavior in the SDQ was calculated (13).

The Short Mood and Feelings Questionnaire (SMFQ), a self-report questionnaire about depression (14, 15), was used to investigate the degree of depression in children in the 1st (10 years old) and 2nd (12 years old) waves of data collection. Each of the 13 items had three response choices: “True” (2 points), “Sometimes true” (1 point), and “Not true” (0 points). The scores for each item were summed into the total score (0–26 points), and higher total scores meant stronger depression.

Previous studies on the relationship between praise for children and prosocial behavior adjusted for the children's sex, age, number of siblings, caregivers' age and educational history (16–18), and one study suggested socioeconomic status as a potential confounder in future studies (18). Therefore, we included children's sex, age, number of siblings, caregivers' age and education, and annual household income as potential candidates for confounders. In addition, using external knowledge, we added children's estimated intelligence quotient (IQ) and psychological distress of primary caregivers (mainly mothers) in the 1st wave as potential candidates for confounders. Children's IQ was estimated from two subsets (Information and Picture Completion) of the Wechsler Intelligence Scale for Children (WISC-III) (19). Psychological distress of primary caregivers was assessed by the Kessler Psychological Distress Scale (K6) (20). Among these potential confounders, we regarded the variables that showed a significant association with the prosocial self-report as confounders.

For comparison of the demographic characteristics between the “prosocial praise group” and the “other praise group,” t-tests, tests of differences in population ratios, or χ² tests were used. To investigate the relationship between being praised for prosocial behavior and depressive symptoms, linear regression analysis was performed using the SMFQ at the 2nd wave (12 years old) as the objective variable and being praised for prosocial behavior as the main explanatory variable. The variables which showed significant difference between the “prosocial praise group” and “other praise group” were treated as confounders. In addition, the SMFQ score at the 1st wave (10 years old) and other confounders were treated as covariates, and multiple regression analysis was performed after supplementing missing values using the multiple substitution method (number of multiple imputations: m = 200). Furthermore, since there is a gender difference in the development of prosocial behaviors in adolescence (18), we examined the interaction effect of sex and prosocial self-reports on depressive symptoms. An interaction term of sex X prosocial self-report was added in the multiple regression analysis. For statistical analyses, open-source statistical software R (version 3.6.1) and the multiple substitution method calculation package mice (version 3.6.0) were used.

Table 1 shows the demographic characteristics of 3,007 pairs of children and primary caregivers who participated in both the 1st (10 years old) and 2nd (12 years old) waves. Regarding the question “What are you praised for?”, 845 (28.1%) children answered that they were praised for prosocial behavior (prosocial praise group), while 2,118 (70.4%) did not report any prosocial behaviors (other praise group). The prosocial praise group had a significantly higher percentage of girls (female: 56.4% vs. 43.7%, p < 0.001) and had more siblings (1.2 ± 0.8 vs. 1.1 ± 0.8, p < 0.05) than the other praise group. Depressive symptoms (SMFQ scores) in the prosocial praise group were significantly lower than those in the other praise group both at the 1st wave (4.3 ± 4.4 vs. 4.9 ± 4.6, p < 0.001) and at the 2nd wave (3.4 ± 4.2 vs. 4.0 ± 4.6, p < 0.01). The rate of children with objective prosocial behavior (subscale score for SDQ) at 10 years old was significantly higher in the prosocial praise group than in the other praise group (7.1 ± 2.0 vs. 6.5 ± 2.0, p < 0.001). The estimated IQ at age 10 tended to be lower in the prosocial praise group (107.2 ± 13.9 vs. 108.0 ± 14.1, p = 0.16). There were no significant differences in the age of children, age of caregivers, psychological distress of primary caregivers (K6 scores), educational background of caregivers, or household income between the two groups.

Table 2 shows the results of simple linear regression analysis and multiple linear regression analyses to investigate the longitudinal relationship between prosocial self-reports and depressive symptoms in children. In the single regression analysis, the children who reported being praised for prosocial behavior at age 10 had significantly lower depressive symptoms at age 12 [partial regression variable: −0.57, 95% confidence interval (CI): −0.96 to −0.17, p < 0.01]. In the multiple regression analysis, in addition to the child's age and sex, objective prosocial behavior, the estimated IQ and the number of siblings were added as confounders because they were significantly different between the prosocial praise group and the other praise group (Table 1). Caregivers' age and education, annual household income, and psychological distress of primary caregiver (K6 scores) were excluded from confounders because there were no significant differences between the prosocial praise and other praise groups. Statistical analysis was performed after missing values were complemented by multiple substitution. Even after adjusting for confounders, being praised for prosocial behaviors at 10 years of age was significantly associated with lower depressive symptoms at 12 years of age (partial regression variable: −0.44, 95% CI: −0.80 to −0.08, p < 0.05). There was no evidence for the interaction effect of sex and prosocial self-report on depressive symptoms (p = 0.22).